Multichannel film dosimetry with nonuniformity correction

Andre Micke,a) David F. Lewis, and Xiang YuInternational Specialty Products, 1361 Alps Road, Wayne, New Jersey 07470

(Received 13 August 2010; revised 24 February 2011; accepted for publication 22 March 2011;published 2 May 2011)

Purpose: A new method to evaluate radiochromic film dosimetry data scanned in multiple colorchannels is presented. This work was undertaken to demonstrate that the multichannel method isfundamentally superior to the traditional single channel method. The multichannel method allowsfor the separation and removal of the nondose-dependent portions of a film image leaving a residualimage that is dependent only on absorbed dose.Methods: Radiochromic films were exposed to 10! 10 cm radiation fields (Co-60 and 6 MV) atdoses up to about 300 cGy. The films were scanned in red–blue–green (RGB) format on a flatbedcolor scanner and measured to build calibration tables relating the absorbed dose to the response ofthe film in each of the color channels. Film images were converted to dose maps using two meth-ods. The first method used the response from a single color channel and the second method usedthe response from all three color channels. The multichannel method allows for the separation ofthe scanned signal into one part that is dose-dependent and another part that is dose-independentand enables the correction of a variety of disturbances in the digitized image including nonuniform-ities in the active coating on the radiochromic film as well as scanner related artifacts. The funda-mental mathematics of the two methods is described and the dose maps calculated from filmimages using the two methods are compared and analyzed.Results: The multichannel dosimetry method was shown to be an effective way to separate out non-dose-dependent abnormalities from radiochromic dosimetry film images. The process was shown toremove disturbances in the scanned images caused by nonhomogeneity of the radiochromic filmand artifacts caused by the scanner and to improve the integrity of the dose information. Multichan-nel dosimetry also reduces random noise in the dose images and mitigates scanner-related artifactssuch as lateral position dependence. In providing an ability to calculate dose maps from data in allthe color channels the multichannel method provides the ability to examine the agreement betweenthe color channels. Furthermore, when using calibration data to convert RGB film images to doseusing the new method, poor correspondence between the dose calculations for the three color chan-nels provides an important indication that the this new technique enables easy indication in case thedose and calibration films are curve mismatched. The method permit compensation for thicknessnonuniformities in the film, increases the signal to noise level, mitigates the lateral dose-depend-ency of flatbed scanners effect of the calculated dose map and extends the evaluable dose range to10 cGy-100 Gy.Conclusions: Multichannel dosimetry with radiochromic film like GafchromicVR EBT2 is shown tohave significant advantages over single channel dosimetry. It is recommended that the dosimetryprotocols described be implemented when using this radiochromic film to ensure the best data in-tegrity and dosimetric accuracy.VC 2011 American Association of Physicists in Medicine.[DOI: 10.1118/1.3576105]

Key words: multichannel, dosimetry, GafchromicVR EBT2, thickness compensation

I. INTRODUCTION

Advanced techniques in radiotherapy, like intensity modu-lated radiotherapy (IMRT) offer dose delivery targeted for

specific lesions. While such advances have provided greatsuccess in reducing dose to healthy tissue and shortened treat-

ment time, they have constantly increased the complexity oftreatment planning and delivery and increased the demand for

quality assurance. Film dosimetry using radiochromic filmlike GafchromicVR EBT is a common tool to verify the dose

distribution of intensity-modulated radiation treatment plansand for general quality assurance of treatment planning sys-

tems and linear accelerators. Due to its high spatial resolution,weak energy dependence1–3 and near-tissue equivalence4

radiochromic film has a suitable characteristics for complextreatment dose verification.5,10 Since GafchromicVR radiochro-mic films produce colored images when exposed to radiation,it has long been known that multichannel flatbed scannersoffer better usability than white-light scanners. A multichan-nel scanner provides a choice of colors for scanning and thishas been useful in offering the selection of the red color chan-nel for greater sensitivity at lower doses, yet being able toextend the dynamic range of the film to much higher dosesusing the green of blue channels.6–9,13

2523 Med. Phys. 38 (5), May 2011 0094-2405/2011/38(5)/2523/12/$30.00 VC 2011 Am. Assoc. Phys. Med. 2523

However, a number of investigators11,12,14,15,21 havepointed to challenges in radiochromic film dosimetry, partic-ularly those related to the scanning of the radiochromic film.To date, radiochromic film dosimetry has relied on the useof information from a single color channel of an red–blue–green (RGB) scanner. The signal thus provided is relativelyweaker at equal dose to that obtained when using a conven-tional silver film. Also, when RGB color scanners are usedto digitize radiochromic film, the images may suffer from anartifact that causes film density values to increase as the lat-eral distance from the scan axis increases.16,19 While theeffect is relatively minor at low doses (<100 cGy) and posi-tions within about 5–7 cm of the scan axis, it can create signif-icant overestimate of higher doses, particularly as filmposition approaches the lateral edges of the scan area. Otherinvestigators have questioned the suitability of EBT2 film fordosimetry application owing to errors from multiple sourcesincluding thickness artifacts in the radiochromic film coat-ing,17 but their dosimetry method utilizes only the data from asingle color channel. The work presented in this paperdescribes a superior approach using data from all the colorchannels to separate the nondose-dependent artifacts due tofilm and scanner from the dosimetric result.

This paper describes a novel approach to film dosimetryusing a colored radiochromic film and an RGB flatbed colorscanner. Since the radiochromic film provides a differentresponse in each of the three color channels, and particularlybecause the slopes of the color response vs dose responsecurves are different in each color channel, this new multi-channel approach to dosimetry enables scanned film imagesto be separated into two parts—one part that is dose-depend-ent and one part that is independent of dose. The dose-inde-pendent part of the image contains information related tothickness, or other response differences in the film coating aswell as to scanner artifacts, including noise, and to someeffects caused by dust particles on the scanner. Having sepa-rated the dose-independent part of the scan information, theremaining dose-dependent information constituting the dosemap has higher fidelity and becomes more useful to the userfor the purposes of the dosimetry.

II. MATERIALS AND METHODS

II.A. Radiochromic film and irradiation procedure

The film used in this study was GafchromicVR EBT2 withsheet dimensions of 20.3! 25.4 cm2. The film was handledaccording to the procedures described in the (AAPM) taskgroup 55 report. Exposure to light was minimized by keep-ing the films in black envelopes when they were not beinghandled for exposure or scanning. Several different sourceswere employed to irradiate the sets of films used for thiswork. This included irradiation with a Co-60 source in anAECL Theratron T-780 as well as irradiation with 6 MVphotons on Varian linear accelerators. For exposures the filmwas placed in a phantom composed of 30! 30 cm sheets ofsolid water or polystyrene with 5 cm of the build-up materialabove and below the film. The source-to-surface distancewas 100 cm. Exposure of film for dose calibration was

performed with 10! 10 cm fields, and the film perpendicularto the axis of the beam. Depending on the source, sets of cal-ibration films were generated using between eight and tendiscrete exposures to doses ranging from about 20 cGy toabout 250 or 300 cGy. One film was used for each doselevel. Measurement and analysis of film sets exposed withthe two radiation sources were kept separate.

Similar solid water or polystyrene phantoms were usedfor the exposure of films to IMRT treatment fields. PatientIMRT films were placed at a depth of 5 cm in the phantomand exposed in clinical mode to the full dose by all fields ofthe treatment plans. Films were scribed with pen-marks priorto irradiation to indicate the positions of the cross-hairs ofthe Linac at 0" gantry angle. Depending on the treatmentplan the maximum doses delivered to the film were in therange from about 200 cGy to about 250 cGy.

II.B. Scanners and scanning

The films were scanned and digitized with an Epsonexpression 10000XL or an Epson V700 flatbed color scan-ner. It is well known that radiochromic films like the EBT2film undergo postexposure changes. Data sets obtained withthe two scanners were kept separate for the subsequent anal-ysis. That is to say the films continue to darken after expo-sure, although the rate diminishes rapidly with time. Toaccount for the postexposure changes all films within a set ofcalibration films were scanned within a time window of lessthan 10 min and at least 24 h after exposure. Under these cir-cumstances errors due to time-after-exposure differences canbe neglected. When IMRT images were obtained for analysisthe images were obtained within 10 min of the calibrationfilm images and at least 24 h after exposure data was col-lected from change in the films. The scanners were fittedwith transparency adapters and the images were acquired intransmission mode. RGB positive images were collected at adepth of 16 bits per color channel and a spatial resolution of72 dpi. Scanning was conducted through the Epson scandriver for each model of scanner. Software settings werechosen to disable all color correction options and therebydeliver the raw scanner data without any photographicenhancements. This choice is critical because it prevents thescan data from being altered to present an image optimizedfor display by adjusting the color balance and exposure.

It is well known that the scan response of EBT2 radiochro-mic is sensitive to the orientation of the film on the scanner.19

Therefore the orientation of the film in each image wasrecorded. In the subsequent measurement and analysis of thecalibration film and IMRT film images care was taken not tomix film images acquired in different orientations. It has alsobeen established that the scanner response of radiochromicfilms like the EBT2 film is sensitive to the position of the filmon the scanner relative to the scan axis. That is the lateralposition on the scanner in the direction perpendicular to thescan direction and relative to the center of the scanner. Thisso called lateral artifact is position dependent and dose-de-pendent. At doses less than about 200 cGy and positionswithin about 5 cm of the scan axis the lateral artifact is less

2524 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2524

Medical Physics, Vol. 38, No. 5, May 2011

than about 2%, but is increasingly important at higher dosesand further away from the central axis of the scanner parallelto the scan direction. To minimize the effect of the lateral arti-fact films were positioned along the central axis for scanning.

One advantage of the triple-channel dosimetry methoddescribed in Subsections II C–II F is that it substantially cor-rects the lateral artifact. To provide data illustrating thisbehavior some scans were made by deliberately positioningthe film to one side of the scan window away from the scanaxis and close to the edge of the window.

II.C. Image measurement and analysis

Scanned images were measured using Film QA Pro soft-ware. Calibration films exposed with 10! 10 cm fields weremeasured by defining an area of interest approximately6! 6 cm in size at the center of the field. Data was obtainedfor the red, green, and blue color channels at a resolution of16 bits per channel. The images are defined as positiveimages where black (no observed signal)—white (maximumsignal) range is mapped to [0, 65 535].

Data and image analysis such as conversion of imagesfrom scanner space to dose space and measurement of filmprofiles and was also performed with the Film QA Pro soft-ware. This application has a utility for calculating doseimages using single, dual, or triple-channel dosimetry asdescribed in the following Subsections II D–II F. It also pro-vides the ability to separate film images into portions thatare either dose-dependent or dose-independent. This is alsodescribed in the following Subsections II E–II F.

Calibration data correlating film response to dose for eachof the three color channels was fit to analytical functionsdescribed as rational functions. In terms of optical density dXat dose D and wavelength X, these functions take the formdx # $ log%%aX & bXD'=%cX & D'' therein aX, bX, and cX arethe constants to be fitted. Rational functions have already thecorrect steady state limit case behavior for very high dosevalues D.

II.D. Single channel film dosimetry method

Radiochromic films, such as GafchromicVR EBT2 developa colored image upon exposure to radiation. The inherentcolor of the image indicates that the optical absorption of theexposed film varies by wavelength.18 The process of dosime-try with such a film involves a number of steps including:exposing the film to radiation; scanning the film to determinethe optical density dX over one or more spectral color bandsof different wavelength X and using the dose-optical densityresponse information for each color channel to convert thescanned image into its dose equivalent (dose mapping).

One of the ways to determine the optical density dX isthrough the use of a color flatbed scanner. This is an imagedigitization device that measures film response over wave-length bands corresponding to the red (R), green (G), andblue (B) bands of the visible spectrum. The scanned opticaldensity is defined as

dX # $ log%X'; (1)

therein Xe [0,1] stands for the normalized color channelstipulated by the output of the scanner. The color channelvalue X depends on scanner coordinates (i,j), i.e., X # Xij

reflects the optical density of the film at coordinates %xi; yj'.The dose mapping process is recognized as a nonlinear

process and its conversion parameters are determined in acalibration process using areas of film (calibration patches)exposed to known doses of radiation. A sufficient number ofhomogeneously exposed calibration patches are scanned togenerate a calibration table fDi; dX%Di'g, l# 1(1)L, where Lis the number of scanned calibration patches.

In accordance with the Beer–Lambert Law, the scannedoptical density value dX%D' at any point is inversely propor-tional to some dimensionless measure s of the thickness ofthe active layer coated on the film

dX%D' # dDX %D's (2)

where the quantity dDX %D' is independent of relative thick-ness s and varies only with the exposure D. It is easy todirectly verify that the model Eq. (2) fulfills the limit cases

lims!0

X%D' # 1 (3)

i.e., film is fully transparent for zero thickness and

lims!1

X%D' # 0 (4)

i.e., film is fully opaque for infinite thickness.When averaging density values dX%D' across the homoge-

neously exposed film region one finds that

dX%D' #1

N

X

i;j

dDX %D's # dDX %D'!s; (5)

where !s stands for the average film thickness with

!s # 1

N

X

i;j

s: (6)

and N is the number of pixels used in the averaging process.The commonly used method to determine a dose value D

from a density value scanned on a film utilizes a single colorchannel X only. The dose is determined as

D # !d$1X dX

s!s

! ": (7)

Assuming that s=s ( 1 , i.e., active layer is perfectly uni-form, Eq. (7) becomes

D # !d$1X %dX': (8)

The calibration function !dX is determined by correlating acalibration table fDi; !dX%D'ig, l# 1(1)L. Members of oneclass of functions used to correlate values in the calibrationtable are known as rational functions. In terms of opticaldensity such functions take the form

!dX%D' # $ loga& bD

c& D

# $; (9)

2525 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2525

Medical Physics, Vol. 38, No. 5, May 2011

where a, b, and c are the equation parameters to be fitted.Figure 1 shows an example of a calibration curve forGafchromicVR EBT2 film as determined from the red colorchannel image from an Epson 10000XL scanner. The calibra-tion table and correlated function parameters are given inTable I. In this example the density values are derived fromthe 16-bit scanner response values, the pixel values PVx%D' ,with x#R, G, B, and the channel values X in Eq. (1) become

X%D' # PVx%D'65535

# $: (10)

The single channel approach is robust in the sense that forany scanned optical density dX a corresponding dose valueD can be calculated. However, if the density value is dis-turbed by the dimensionless amount DdX (e.g., because ofvariation of the thickness s of the film coating or because ofa scanner nonlinearity like the lateral distortion) oneobtains a corresponding disturbance DD of the dose valuewith

D%1& DD' # !d$1X %dX%1& DdX''; (11)

This appears directly as error of the dose value D and pro-vides no indication of the discrepancy.

II.E. Dual channel film dosimetry method

From Eq. (7) it is clear-cut that picking two scanningwave lengths, i.e., two color channels X1 and X2, results in asystem of two equations and two variables: dose D and rela-tive thickness s=s s

!s .

D # !d$1X1

dX1

s!s

! "

D # !d$1X2

dX2

s!s

! " (12)

and requires the calibration functions !dX1and !dX2

of theselected color channels. This equation system can be sepa-rated to obtain

D # !d$1X1

dX1

dX$1

!dX2%D'

# $(13)

FIG. 1. Single channel calibration curve and data flow of dose mapping using Eq. (8).

TABLE I. EBT2 @ Epson 10000XL calibration table and correlated function

parameters for R, G, and B.

D=cGy dR dG dB

250 0.31776 0.26153 0.56340

225 0.30324 0.25026 0.55564

200 0.28148 0.23533 0.54572

175 0.26510 0.22176 0.53753

150 0.24328 0.20905 0.53208

125 0.22144 0.19456 0.52267

100 0.19799 0.17914 0.51279

75 0.17431 0.16398 0.50556

50 0.14685 0.14811 0.49630

25 0.11521 0.13128 0.48812

0 0.08137 0.11710 0.47790

aX 2.22726 5.22144 3.01197

bX 0.10730 $0.05009 0.06079

cX 2.68937 6.81595 9.04245

2526 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2526

Medical Physics, Vol. 38, No. 5, May 2011

and

s!s#

!d%D'X1=2

dX1=2

(14)

which requires only the solution of a single nonlinear equa-tion and evaluation of the explicit Eq. (14) for one of thecolor channels employed.

This method works optimally for low exposures of EBT2film by selecting X1 # R and X2 # B and using the bluechannel to evaluate Eq. (14). If the exposure is low enoughthe density recorded in the blue channel becomes insensitiveto dose, i.e., !dB%D' ) !d

0B # !dB%0', and Eq. (13) becomes

D # !d$1R

!d0B

dRdB

# $# !d

$1R=B

dRdB

# $(15)

which calculates the dose D as a function of the ratio of theresponses in the red and blue color channels. The foregoingis equivalent to calibrating dose D vs color channel ratiodR=dB # 1 dR=dB as proposed in Ref. 8 p.10, as referencechannel method.

Note that if the dose value D is known (using any proce-dure), it is still possible to use Eq. (14) to calculate the rela-tive layer thickness s=s using the calibration function of anadditional channel. However, the limiting single channeldosimetry case results in s=s # 1, and it is not possibleto separate the dose map into its dose-dependent and dose-independent parts.

The dual channel dose mapping can account for signaldisturbances caused by thickness variations and the like, butthe results may depend on the exposure range used. Equation(13) is not symmetric with respect to the chosen color chan-nels. The method will work best for low exposures of EBT2film less than about 10 Gy when X1 # R and X2 # B, and forhigh exposures greater than about 30 Gy when X1 # B andX2 # R. In the latter case the film response in the red channelapproaches saturation and the term !dR is dominated by thevalue of the relative layer thickness s=s in the same way asthe blue channel response is dominated at low exposures dueto the presence of the yellow marker dye. The dual channelmethod uses explicitly the Beer–Lambert Law in form of Eq.(2) to describe the disturbance Dd, i.e., this method is notable to remove disturbances that are caused by effects otherthan thickness variation of the active film layer. However,the use of Eq. (2) allows to separate the equation system andreduces the calculation of the dose D to a simple inversionof the single nonlinear Eq. (13).

Since the dual channel method is a special case of themultichannel method described in the following SubsectionsII F, specific examples are not presented.

II.F. Triple-channel film dosimetry method

The dual channel method addresses only disturbancescaused by thickness variations of the active film layer. Themore general target is to separate numerically the dose-de-pendent part of a scanned optical density signal from anydisturbance Dd that might be present in the system consist-

ing of radiation source, radiochromic film, and film scannerunit. To model this case, Eq. (2) is modified to

dX%D' # dDX %D'Dd (16)

which includes the previous case for Dd #s=s. Note thatwhen calibration films are used to determine !dX, it is impor-tant that the averaging regions, i.e., the exposed measure-ment areas, must be sufficiently large to ensure that thecalibration condition for the average disturbance

D!d # 1 (17)

is fulfilled. This is equivalent to requesting that the measure-ment area should be large enough to reflect average responseof the system, otherwise !dX will develop a systematic bias.The disturbance Dd is dose-independent since all dose-de-pendent parts are presented by the calibration function !dX.

A dose value can be calculated for each color channel Xusing (Fig. 2)

DX # !d$1X %dXDd': (18)

Since the dose cannot depend on the color channel Xselected for evaluation of Eq. (18). one can consider asequence of multiple channels fXkgKk$1 and minimize the dif-ferences in the dose results from the individual color chan-nels, i.e.,

X%Dd' #X

i 6#j

%DXi $ Dxj'2 ! min

Dd: (19)

The solution to this least square equation can be found bysolving the single nonlinear equation

d

dDdX # 0: (20)

This minimization is equivalent to finding the “nearest”color to the color path f!dR%D'; !dG; !dB*%D'g. The distance of apoint from this color path represents the disturbance value,Dd, and the value of the path parameter is the “best” dosevalue. This method works as long as the slopes of the cali-bration functions !dx for the individual color channels are suf-ficiently different. The more color channels involved, themore deterministic the system becomes, and the better willthis method separate the dose-dependent part of the signalfrom the dose-independent part.

In case of two color channels (K# 2) Eq. (20) becomesDX1

# DX2# 0 and with D # DX1

# DX2and s=s # Dd one

recovers Eq. (12) of the dual channel method.

III. RESULTS

III.A. Single channel film dosimetry

Figure 3 shows the scanned image of an EBT2 filmexposed to a dose of 225 cGy using a 10! 10 cm2 flat fieldexposure and the corresponding dose map based on Eq. (8)for X#R, G, B.

Nonuniformities in the form of vertical stripes areobvious by visual inspection of both the film image and thedose map. A horizontal profile across the dose map in Fig. 4reveals the oscillations especially in the flat part of the

2527 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2527

Medical Physics, Vol. 38, No. 5, May 2011

exposure field, which must stem from dose disturbances DDdescribed in Eq. (11).

Figure 4 shows the dose profiles across the flat field ascalculated separately for the R, G, and B color channels. TheEBT2 film includes a yellow marker dye in the active coat-ing8 that significantly increases the dependence of the signalin the blue color channel with respect to disturbances in thethickness of the active layer. The profile of the blue channelin Fig. 4 suggests that the disturbances DD are dominated bythe variations of the term s=s, i.e., the nonuniformity of theactive layer in the film causes this type of oscillation andwhen using a single channel based dose mapping method,results in dose error.

In the following Subsections III B–III D dose mappingmethods are presented that account for thickness variations

of the active film layer, separate these variations from thedose-dependent portion of the image and remove their effectfrom calculated dose maps.

III.B. Triple-channel film dosimetry

Figure 2 shows results for the same EBT2 films as given inFig. 1 and Table I except that calculations have been madeusing the multichannel method. Compared to the single chan-nel data approach, the multiple channel method varies thedose values until the corresponding scanned optical densityvalues for the various color channels are best matched.

After extracting the disturbance value Dd from Eq. (20),one can calculate the corresponding dose values DX for eachcolor channel. Figure 5 shows the dose map fDR;DG;DBg,and the disturbance map using Eq. (14) with DZ, X#R, G,and B for the scan depicted in Fig. 3. The separation of thedose-dependent part and dose-independent parts are recog-nizable. Whereas the dose map represents the intrinsic resultof the dosimetric measurement, the disturbance map givesinformation about the error sources of measurement process.

The horizontal profile of the dose map in Fig. 6 shows theimproved flatness across the exposed area for all involvedcolor channels. The corresponding profile across the disturb-ance map consists of a low frequency pattern dominated bythickness variations of the active layer (see also s=s). Thepattern becomes more visible by applying a simple averagefilter to the disturbance map (see right part of Fig. 7). Thedisturbance signal in the example varies by a little more than61% around the average value of 100% as defined by thecalibration patches. The latter figure is consistent with the

FIG. 3. EBT2 film with 225 cGy exposed and corresponding dose mapimage showing nonuniformities stemming from thickness variations of theactive film layer [Calculations were carried out using ISP software packageFilm QA Pro 2010 (Ref. 20)].

FIG. 2. Multichannel calibration curves and data flow of dose mapping using Eq. (8).

2528 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2528

Medical Physics, Vol. 38, No. 5, May 2011

aim of having the size of the calibration patches largeenough to represent the average system response includingthe film and the average thickness of the active layer.

Since nonuniformities are broadly distributed in the activelayer of the film, it is good practice to expose and measurelarge calibration patches. This makes it more probable thateach of the measurements will capture the average responseof the film. If small pieces of film are used for the calibration,it is more likely that they will have a response significantlyabove or below the average. As a practical matter, it is pre-ferred to use the calibration patches sized at least 5! 5 cm2

and even more desirably, 10! 10 cm2. Because larger calibra-tion patches better represent average film behavior, a smallernumber are necessary to construct the response curve. Experi-ence shows that a calibration for a dose range from 0 to300 cGy is better served by using six to eight 10! 10 cmpatches rather than two to three times that number of smallerpatches. Whereas the single channel dosimetry method isindiscriminate and converts any thickness variations directlyinto erroneous dose values (see Ref. 8), the multichannelmethod described accounts for such disturbances, resolves

them from the dose information and places them into a sepa-rate map displaying the relative thickness deviations.

III.C. Comparing dosimetry methods using IMRT dosedistributions

Figure 8 displays the dose plan image in the coronal planeof an IMRT treatment composed of seven beams and the cor-responding dose image recorded on EBT2 radiochromic film.The radiochromic IMRT film and eight calibration filmsexposed to 6MV x-ray doses between zero and 281 cGy weredigitized in transmission mode on an Epson 10000XL scan-ner. The dose response is shown in Fig. 9 where the calibra-tion data for each color channel have been fit to a function ofthe type X(D)# (a& b D)=(c&D), where the scannerresponse at dose D is X(D) and a, b, and c are constants.

Using this calibration data dose images were calculatedfrom the IMRT measurement film image using the single, redchannel dosimetry method, and the triple-channel methodusing the red, green, and blue response data. Isodose maps arepresented in Fig. 10 showing the agreement between dosecontours in the treatment plan and the measured dose contoursin the single-channel triple-channel dose maps. Contour linesare shown for doses at 90, 70, 50, and 30% of the maximumvalue in the plan map. Inspection of the iso-contour mapsshows that the contours in the triple-channel map have acloser conformance to the plan than those in the single chan-nel map. Quantitative assessment of the accordance betweenmeasurements and plan was made using the gamma functiondistribution using factors of 3% dose agreement within 2 mm.The results are shown in Fig. 11 in the form of a gamma func-tion histogram. While there is measured agreement of 91%between the single channel dose map and the plan the triple-channel dose map has a 99% agreement, a much closer con-cordance with the plan. The measurements and comparisonswere repeated for exposures of four other EBT2 films to thesame IMRT treatment plan. Dose maps were calculated using

FIG. 4. Horizontal profile across the dose map for 25 and 225 cGy (see Fig. 3) exposure showing red, green, and blue channel results.

FIG. 5. Dose map and disturbance map for EBT2 film with 225 cGy expo-sure shown in Fig 3.

2529 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2529

Medical Physics, Vol. 38, No. 5, May 2011

the single (red) channel and triple-channel methods. Theagreements with plan using the single channel dosimetry pro-tocol were 90.4, 86.0, 98.1, and 83.3 (average 89.5) for thefour measurement films. For maps calculated using the triple-channel method the agreements were 99.6, 99.2, 99.8, and98.4 (average 99.2). Radiochromic film dosimetry with EBT2film consistently provided better concordance with the treat-ment plan when the triple-channel protocol dosimetry wasused rather than the single channel method.

III.D. Mitigation of lateral displacement effect

Because EBT2 film partially polarizes transmitted light,22

the doses calculated using images from a CCD scanner mayshow substantial lateral distortions.16,20,21,23 The multichan-nel technique is able to mitigate this lateral dependence. To

demonstrate this, one of the EBT2 films exposed to a10 m! 10 cm field and used to generate the calibration curveshown in Table I was scanned in the center of an Epson10000XL scanner and then moved to a new position dis-placed “4” laterally from the center of the scanner andrescanned a position close to the left edge of the scan win-dow. The calibration fitting functions were used to convertthe images to dose maps using the single channel and triple-channel mapping. Profiles in the lateral direction across theexposed areas of the film scanned close to the edge of thescan window are shown in Fig. 12. The left side of each pro-file is furthest from the scan axis. The single channel profileshows a large increase in the dose on the left side of the field,i.e., the side of the field farthest from the center of the scan-ner. Qualitatively this behavior is characteristic of theresponse of CCD scanners16,18 including those made by

FIG. 6. Horizontal profile across the dose map for 25 and 225 cGy exposure (see Fig. 5) showing the red, green, and blue channel results.

FIG. 7. Horizontal profile across the dose map from Fig. 5 for red, green, and blue channel.

2530 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2530

Medical Physics, Vol. 38, No. 5, May 2011

Epson, Microtek, and Vidar. The indicated optical densitytends to be greater approaching the lateral edges of thescanner’s field of view. Not only is this effect sensitive toposition on the scanner, but it is also dose-dependent, beingsmaller at low doses, but becoming more substantial athigher doses. The triple-channel method will compensatefor the dose-independent part of the lateral effect (e.g., atzero exposure). However, the dose-dependent part willremain.

By comparison, the profile of the same area of the triple-channel map shows that the apparent increase in dose closerto the edge of the scanner can be mitigated by using all threecolor channels to perform the dosimetry as depicted inFig. 12. This is further demonstrated by showing the profileacross the triple-channel map of the laterally displaced filmtogether with the profile across the same area of the samefilm scanned at the center of the scanner. These profiles aredemonstrating how the triple-channel dosimetry method is

FIG. 8. Dose images calculated for IMRT plan (right) and measured with EBT2 film.

FIG. 9. Dose calibration curves for EBT2 film in three color channels.

2531 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2531

Medical Physics, Vol. 38, No. 5, May 2011

mitigating the lateral scanner effect and delivers resultssuperior to the single channel method.

The dose values calculated for the different color chan-nels with the multichannel method should be in close agree-ment. Offsets indicate that the dose-dependent properties ofthe scanned film do not match sufficiently with those of thecalibration films or the calibration condition (17) is not

fulfilled. Comparison of the dose values from the color chan-nels makes it possible to pinpoint local defects and errors ofthe scanned radiochromic film and the scanning protocol.

The example given in Fig. 13 is based on the scan of the200 cGy calibration patch used in Table I, but in this case thefilm was rotated by 90" before scanning. It is well known thatthe response of EBT2 film is dependent on its orientation on

FIG. 10. Iso-dose maps– measurement (thin lines) vs IMRT plan (thick lines): Red channel dosimetry to left; triple-channel dosimetry to right.

FIG. 11. Gamma distribution and histogram maps: Red channel dosimetry, 91.0% agreement, to left; triple-channel dosimetry, 99.0% agreement, to right.

2532 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2532

Medical Physics, Vol. 38, No. 5, May 2011

the scanner and the calibration Table I obtained for one orien-tation will not match the film behavior in another orientation.When dosimetry is performed using the single channelmethod, there is no indication in the dose result to signify that

a film was misaligned when scanned. However, such an indi-cation is inherent when using multichannel dosimetry becausethe doses from the color channels should be in close agree-ment. In the example, there is an offset of about 10 cGy

FIG. 12. Comparison single and triple channel method using dose map from film positioned at left side of the scanner where lateral effect reaches maximumand single and triple channel dose map of film at the center of the scanner.

FIG. 13. Horizontal profile across the dose maps of 200 cGy film patch scanned rotated by 90" compared to calibration scans (left) and the scan with originalorientation (right) for R, G, B channels.

2533 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2533

Medical Physics, Vol. 38, No. 5, May 2011

(approximately 5%) between the color channels, providing aclear indication of a mismatch between the scanned film andcalibration data used to calculate the dose map.

The multichannel method is symmetric with respect to allcolor channels used in the measurement. It characteristicallybalances the color channel with the highest sensitivity becausethe component with the highest derivative value is the domi-nant factor in Eq. (18). The use of multiple wavelengths toscan the absorption spectrum of the film makes it possible touse the entire sensitivity range of the film covered by the spec-tral response of the scanner. This range could even beextended by using additional wave lengths or a larger numberof color bands, although this would require the utilization of aspecialized scanner. In the case of using EBT2 film and Epsonflatbed color scanners (e.g., models 10000XL, V700, V750,1680, and 4900) the RGB channels cover a dose measurementrange from below 10 cGy to above 100 Gy, i.e., more than 3orders of magnitude (see Ref. 8).

IV. CONCLUSIONS

The fundamentals of a new method using multiple colorchannels to convert scanned images of radiochromic filminto dose maps have been presented. This method allows theseparation of the dose-dependent and dose-independent partsof a scanned signal and compensates for a variety of anoma-lies, artifacts, and other disturbances, such as variations ofthe thickness of the active layer, scanner nonlinearity andprocess noise and enables the use of the entire available sen-sitivity range of the film in the same procedure. The substan-tial gain in accuracy is illustrated by improved flatness andsymmetry of the tested flat field exposures.

NOMENCLATURE

Symbol(SI unit) ExplanationD (Gy) # absorbed dose

X # normalized color channel value, Xe [0, 1],typical color channels are R, G, B

R, G, B # normalized of red, green, blue colorchannel, Xe [0, 1]

dX # optical density of color channel X,dX # $ log%X'

dDX # dose-dependent part of the optical den-sity dX

!dot # script to mark average value of a quantitys # relative thickness of the active film layer

Dd # disturbance factor of optical density valuex, y (m) # absolute locations x and y coordinates

i, j # discretized locations in x (i) and y (j)direction of pixelated coordinate system

ACKNOWLEDGMENTS

The authors are employees of International Specialty Prod-ucts the manufacturer of GafchromicVR dosimetry film andowner of the FilmQA Pro software (www.FilmQA Pro.com).

a)Author to whom correspondence should be addressed. Electronic mail:amicke@ispcorp.com

1M. J. Butson, T. Cheung, and P. K. Yu, “Weak energy dependence of EBTGafchromic film dose response in the 50 kVp–10 MVp x-ray range,”Appl. Radiat. Isot. 64, 60–62 (2006).

2S. T. Chiu-Tsao, Y. Ho, R. Schankar, L. Wang, and L. B. Harrison,“Energy dependence of response of new high sensitivity radiochromicfilms for megavoltage and kilovoltage radiation energies,” Med. Phys. 32,3350–3354 (2005).

3P. Lindsay, A. Rink, M. Ruschin, and D. Jaffray, “Investigation of energydependence of EBT and EBT-2 Gafchromic film,” Med. Phys. 37(2), 571–576 (2010).

4A. Niroomand-Rad, C. R. Blackwell, B. M. Coursey, K. P. Gall, J. M.Galvin, W. L. McLaughlin, A. S Meigooni, R. Nath, J. E. Rodger, andC. G. Soares, “Radiochromic film dosimetry: Recommendations of AAPMradiation therapy committee task group 55, Med. Phys. 25, 2093–2115(1998).

5AAPM Report No. 63, Radiochromic film dosimetry, 1998.6S. Devic, J. Seuntjens, G. Hegyi, E. Podgorsak, C. Soares, A. Kirov, I. Ali,J. Williamson, and A. Elizondo, “Dosimetric properties of improved Gaf-chromic films for seven different digitizers,” Med. Phys. 31(9), 2391–2398 (2004).

7S. Devic, N. Tomic, C. Soares, and E. Podgorsak, “Optimizing thedynamic range extension of a radiochromic film dosimetry system,” Med.Phys. 36(2), 429–437(2009).

8GafchromicVR EBT2 Self-Developing Film for Radiotherapy Dosimetry,ISP White Paper, October, 2010.

9E. Willcox, G. Daskalov, and L. Nedialkova, “Comparison of the Epsonexpression 1680 flatbed and the VXR-16 dosimetry PRO film scanners foruse in IMRT dosimetry using Gafchromic and radiographic film,” Med.Phys. 34(1), 41–48 (2007).

10M. Fuss, E. Sturtewagen, C. De Wagter, and D. Georg, “Dosimetric char-acterization of GafChromic EBT film and its implication on film dosime-try quality assurance”, Phys. Med. Biol. 52(14), 4211–4225 (2007).

11S. Devic, J. Seuntjens, and E. Sham, “Precise radiochromic film dosi-metry using a flat-bed document scanner,” Med. Phys. 32(7), 2245–2253(2005).

12B. D. Lynch, J. Kozelka, and M. K. Ranade, “Important considerations forradiochromic film dosimetry with flatbed CCD scanners and EBT Gaf-chromic Film,” Med. Phys. 27(10), 2462–2475 (2006).

13ISP presentation: www.FilmQA Pro.com/Micke_Lewis_Yu_Triple_Channel_Technique_201001.pdf.

14L. Paelinck, W. de Neve, and C. De Wagter, “Precautions and strategies inusing a commercial flatbed scanner for radiochromic film dosimetry,”Phys. Med. Biol. 52(1), 231–242 (2007).

15L. J. Van Battum, D. Hoffmans, D. Piersma, and S. Heukelom, “Accuratedosimetry with Gafchromic EBT film of a 6 MV photon beam in water:What level is achievable?,” Med. Phys. 35(2), 704–716 (2008).

16J. Menegotti, A. Delana, and A. Martignano, “Radiochromic film dosime-try with flatbed scanners: A fast and accurate method for dose calibrationand uniformity correction with single film exposure,” Med. Phys. 35(7),3078–3085 (2008).

17B. Hartmann, M. Martisikova, and O. Jakel, “Homogeneity of GafchromicEBT2 Film,” Med. Phys. 37(4), 1753–1756 (2010).

18S. Devic, S. Aldelaijan, H. Mohammed, N. Tomic, L. H. Laing, F.DeBlois, and J. Seuntjens, “Absorption spectra time evolution of EBT-2model Gafchromic film,” Med. Phys. 37(5), 2207–2214 (2010).

19L. Paelinck, A. Ebongue, W. De Neve, and C. De Wagter, “RadiochromicEBT film dosimetry: Effect of film orientation and batch on the lateral cor-rection of the scanner,” Radiother. Oncol. 84, 194–195 (2007).

20FilmQA Pro 2010 software: www. FilmQA Pro.com.21L. J. Van Battum, D. Hoffmans, S. Kwa, and S. Heukelom, Accuracy ofGafChromic EBT Film as dose meter in radiotherapy QA, Proc. IFMBE,25, 105–108 (2009).

22M. J. Butson, T. Chueng, and P. K. Yu, “Evaluation of the magnitude ofEBT Gafchromic film polarization effects,” Australas. Phys. Eng. Sci.Med. 32(1), 21–25 (2009).

23S. Saur and J. Frengen, “GafChromic EBT film dosimetry with flatbedCCD scanner: A novel background correction method and full dose uncer-tainty analysis,” Med. Phys. 35(7), 3094–3101 (2008).

2534 Micke, Lewis, and Yu: Multichannel film dosimetry with nonuniformity correction 2534

Medical Physics, Vol. 38, No. 5, May 2011